Incorporating prior knowledge into multi-label boosting for cross-modal image annotation and retrieval

Abstract

Automatic image annotation (AIA) has proved to be an effective and promising solution to automatically deduce the high-level semantics from low-level visual features. In this paper, we formulate the task of image annotation as a multi-label, multi class semantic image classification problem and propose a simple yet effective joint classification framework in which probabilistic multi-label boosting and contextual semantic constraints are integrated seamlessly. We conducted experiments on a medium-sized image collection including about 5000 images from Corel Stock Photo CDs. The experimental results demonstrated that the annotation performance of our proposed method is comparable to state-of-the-art approaches, showing the effectiveness and feasibility of the proposed unified framework. © Springer-Verlag Berlin Heidelberg 2006.